Handwriting recognition software has made it possible to digitally capture handwriting and transform it into digital characters using an input capture device and a computer. The capture device may be a flat panel device that allows a user to enter normal handwritten notes onto a piece of paper attached to the capture device while information about the coordinates of the pen strokes is digitally recorded by the capture device. The capture device can later upload the digitally recorded handwritten notes into a computer where an uploading program receives and stores the handwriting strokes in memory, resulting in two copies of a document, namely the original handwritten version and a second, digitally encoded version.
Digital handwriting capture is useful when data must be entered into a computer program for processing, but original handwritten copies must be retained for legal or verification purposes. In these instances, it would be helpful to have handwriting automatically transformed into digital characters and transferred to a computer program without manual data entry. This may be achieved by placing a printed paper form with clearly defined input fields on a capture device, digitally capturing the handwritten notes in these input fields on the capture device, and uploading the digital notes to the computer. A recognition program may then interpret the digitally recorded handwritten notes within these input fields and transform them into a digitally encoded representation, which can be automatically entered into the computer program in the same manner as if the notes were manually entered via a keyboard.
However, the problem with these systems is that handwriting recognition software is generally not 100% accurate. Conventionally, an accuracy of approximately 90% can be reached, but only after a user has gone through a series of arduous user-specific training sessions that allow the recognition software to adjust to the style of a particular user. So, time is lost by having to train the software. But, even so, there may still be multiple transformation errors because of the margin for error in current handwriting recognition software.
Errors are particularly prevalent when a user enters calendar date information in numerical form. For example, if an input field is provided for date entry and no format is given, a user may enter the date in one of many formats. Dec. 2, 1975 may appear as 12.02.1975 or 12/02/75 in the U.S. and 02.12.1975 or 02/12/75 in Europe, for example. The ambiguous nature of the date field results in difficulty in resolving handwritten input as a date understood by the computer program that will later use this information. Additionally, even if separate fields are provided for day, month, and year and the expected format is clearly marked on the printed form, users may still unintentionally enter the date in an incorrect format if they ignore this additional instruction.
Finally, even if users correctly follow the format required, other difficulties may arise when resolving individual numbers. Different cultures may write numbers differently. For example, the handwritten German 1 closely resembles a 7 and is interpreted as such by handwriting recognition software developed in the U.S. Or handwriting recognition engines may lack robustness in interpreting numbers. For example, handwriting recognition engines that are trained on a specific user's input may not correctly interpret hastily written numbers by the specific user and can not at all interpret numbers from a different user. In some cases, the number 8 could easily be interpreted as the letter B.
Some systems have tried to solve these problems with date recognition by providing graphical user interfaces through which a user may select date information from pull-down menus. In these systems, a more complex input/output device than the capture device must be used to display the graphical user interfaces. Such a device could be expensive and too bulky to carry, particularly for field surveys, field inventory, etc., for which the capture device is ideally suited.
Accordingly, there is a need in the art for a simple and natural way to improve the recognition accuracy of calendar date information entered by a user onto printed paper forms attached to capture devices independent of the user who inputs the information.